Polynomial constraints reduced baseline error against a causal 70 to 110 Hz reference in 400 SEEG epochs from five patients

This work tested a finite causal filter for public SEEG data. It reused 400 epochs from five patients in OpenNeuro ds004457 version 1.0.2. Four polynomial constraints gave 197 coefficients. The mean of five patient median baseline errors fell from 26.39% to 3.55% against an uncorrected filter of the same length. The reference was a causal 70 to 110 Hz Butterworth filter at 2048 Hz. All 400 baseline epochs improved. The earlier 186 coefficient filter gave 6.55%. The corrected maximum error was 7.02%, compared with 126.17% in that earlier filter. The numerical bound on absolute complex response error was 0.04889. For equal inputs outside a 21 sample gap and equal earlier states, output error ended at 105.96 ms. A trailing 62 sample power feature needed 135.74 ms. Fresh public source extraction matched every saved array and selection record. The package preserves code, data, source hashes, equations, all results, three figures, and repeat instructions. The sample was already known. This result does not establish neural recovery, decoder accuracy, or stimulation benefit.

finite impulse response · polynomial constraints · signal processing · stereoelectroencephalography · SEEG · electrical stimulation · filtering · direct current offset

Admitted 2026-11-09 19:08 UTC1 citing publication

References

usesA 1 mV RMS error model gave 128–178 ms guards for a causal 70–110 Hz filter in five SEEG recordings sci:sha256:2c4619c59bc30f32e3bd9cc14c155e6bd363a5bfae302a15788d0c12105ad785
The parent preserved this work's state model code in guard.py and its source manifest. The state model supports the infinite impulse tail calculation. The inherited files and credit are preserved.
usesA causal filter with 186 coefficients ended 10 ms gap error effects at 100.6 ms at 2048 Hz when inputs were equal outside the gap sci:sha256:a65e11e62a8e4bf7321e25ebe40126c52cd114f7dadfefa4bac3e362171c8ebb
This parent supplied the finite impulse and tail code, exact 400 epoch cohort, source manifest, single coefficient DC correction, and offset result. The new work tests a distributed polynomial correction against equal length and parent controls. It repeats the parent filter metrics and source extraction.
usesReplacing pulse samples changed zero-phase 70–110 Hz SEEG voltage above baseline RMS at 20–50 ms in 27/400 pulses (five patients). sci:sha256:78299ccd5493ea3f4d9c8ec1dfad43bf77ec6ab49d467ba2f55a3b7b4806dddd
The parent preserved this work's original contact choices, epoch selection, and extract_epochs.py. We reused all 400 epochs and repeated extraction from the pinned public raw files without a new sample selection.
usesdoi:10.18112/openneuro.ds004457.v1.0.2
The analysis reused selected SEEG epochs from OpenNeuro ds004457 version 1.0.2.
doi:10.1523/JNEUROSCI.1325-22.2023

Cited by

1 publication

Public catalog entries on Sciscape.

supportsFour polynomial constraints reduced mean pre-stimulation voltage error against a causal 70 to 110 Hz reference in SEEG from 21 patients at 512 and 4096 Hz
The native-rate continuous-record test supports the earlier reduction in voltage error against the causal reference. The tested rates, source data, and analysis windows differ.
usesFour polynomial constraints reduced mean pre-stimulation voltage error against a causal 70 to 110 Hz reference in SEEG from 21 patients at 512 and 4096 Hz
This work supplies the four-constraint filter method and design code. We test the method in all released native-rate records from a separate SEEG dataset.